Parameterization of Cumulus Convective Cloud Systems in Mesoscale Forecast Models

Transcription

1 DISTRIBUTION STATEMENT A. Approved for public release; distribution is unlimited. Parameterization of Cumulus Convective Cloud Systems in Mesoscale Forecast Models Yefim L. Kogan Cooperative Institute for Mesoscale Meteorological Studies University of Oklahoma phone: (405) ; fax: (405) ; Grant #: N LONG-TERM GOALS The development and improvement of the microphysics parameterization of cumulus convective clouds in mesoscale numerical weather prediction models OBJECTIVES Conduct detailed studies of cloud microphysical processes in order to develop a unified parameterization of boundary layer stratocumulus and trade wind cumulus convective clouds. Develop a parameterization of subgrid cloud variability that in combination with the unified parameterization of conversion/sedimentation rates will provide a complete formulation of microphysical processes in convective clouds for use in mesoscale forecast models. Test the parameterization using COAMPS model in simulations of convective cloud systems. APPROACH The research is based on the SAMEX large eddy simulation (LES) model with explicit formulation of aerosol and drop size-resolving microphysics. The LES simulations based on observations from field projects will provide datasets necessary for parameterization development. COAMPS simulations based on field projects data will test the parameterization. WORK COMPLETED The following tasks have been completed during the 3 rd year of the project: 1. In the previous years we developed a probability distribution function (PDF) formulation for shallow cumulus clouds to allow application of our cloud microphysics parameterization in mesoscale prediction models. We now have expanded the PDF formulation for use in cumulus congestus clouds. The cumulus congestus parameterization was developed based on the data from the TOGA COARE field campaign. The formulation of the parameterization based on fully twodimensional joint PDFs represents most accurately autoconversion and accretion rates on a mesoscale model grid. 1

2 2. We completed the sensitivity experiments to investigate the possibility of using a single universal PDF formulation for both shallow and congestus clouds. Additional sensitivity experiments were conducted to investigate possibility of using universal PDFs in NWP models with different horizontal grid resolutions. RESULTS 1. A PDF based microphysics parameterization for congestus cumulus clouds Marine trade cumulus are one of the most prevalent cloud types on Earth and play an important role in establishing the thermodynamic structure of the lower atmosphere in the trade latitudes. Satellite observations from the Tropical Rainfall Measuring Mission (NASA TRMM) project found that the shallow and congestus modes together contribute about 20% of the total tropical oceanic precipitation, with the bulk of the rainfall coming from the deeper congestus clouds. Accurate forecast of precipitation requires unbiased calculations of microphysical process rates such as autoconversion and accretion, which in a mesoscale NWP model necessitates accounting for SGS variability over the model grid volume. This variability can be expressed as probability distribution functions (PDFs) of microphysical variables, and the process rates can be integrated over these PDFs in order to obtain the unbiased grid-averaged process rates. Using dynamically balanced LES results from a simulation of a marine trade cumulus cloud system initialized with TOGA COARE soundings, we obtained 1D PDFs of the individual variables q c, N c, and q r, as well as joint 2D PDFs (JPDFs) of pairs of variables (q c, N c ) and (q c, q r ); the latter are used to calculate, respectively, autoconversion and accretion in bulk models. Both 1D PDFs and 2D JPDFs are best formulated as functions of their non-dimensional parameters, i.e. normalized using the layer-mean parameter values. Such formulation exhibits weaker height dependence and, hence, allows for more robust parameterization of PDF parameters. The 1D PDFs can be approximated by analytical functions, however, in a more complex form and with stronger variation in the vertical. The 1D PDFs of droplet concentration N c can be fitted by a Gaussian distribution, while the cloud water and rain water mixing ratio distributions are highly skewed and can be represented by lognormal distributions. As in the case of shallow cumulus clouds, we find that JPDFs of q c and N c (inputs to the autoconversion) cannot be represented by a simple product of their individual 1D PDF; only the exact JPDF captures correctly the covariation between q c and N c. The JPDFs of q c and q r (inputs to the accretion rate formula), on the other hand, can be approximated more accurately by the product of the individual 1D PDFs of q c and q r. The process rates calculated using different formulations of the PDFs are compared to the benchmark process rate calculated using the point-by-point variability over the LES grid. Specifically, we calculate the bias in autoconversion and accretion rate. Using the JPDFs results in the most accurate values for both autoconversion and accretion (Figure 1). Approximating the 2D JPDF by using a product of individual 1D PDFs overestimates the autoconversion rates by more than an order of magnitude, whereas neglecting the SGS variability altogether results in a drastic underestimation of the grid-mean autoconversion rate. Differences in accretion rate for different PDF assumptions are smaller compared to autoconversion rate, largely a result of the weaker correlation between variables q c and q r and the near-linear dependence on the variables in the accretion rate formula. 2

3 Figure 1. Comparison of errors using various formulations of autoconversion (left panels) and accretion (right panels) conversion rates. Top panels show scatter plots of parameterized conversion rates vs the benchmark conversion rates derived from the point by point averaged rates. Bottom panels show cumulative distribution of errors using different formulation of PDFs (black symbols and curves the new 2D formulation, red the 1D formulation, blue the formulation with no PDFs and using only the mean values, and magenta the simplified version of the 2D formulation using a fixed time PDF.[graph: The joint fully 2D PDFs approximate conversion rates most accurately] Because of the impracticality of using PDFs that are evolving in time during the mesoscale model simulation, we evaluated an additional PDF approximation, which is fixed in time but height dependent. Results (shown in bottom panels in Fig. 1 by the magenta curves) suggest that use of a single (fixed in time but height dependent) JPDF for each pair of variables gives an acceptable level of precision which is only slightly less accurate than the time-dependent PDFs. 2. Can universal PDFs be used for both shallow and congestus clouds, as well as for all NWP model grid resolutions? We conducted the sensitivity experiments to investigate the possibility of using a single universal PDF formulation for both shallow and congestus clouds. Results shown in Figure 2 demonstrate that accretion rates can be approximated with about the same level of accuracy using a universal PDF. However, the errors in autoconversion rates are much more sensitive to the tuning of the PDFs to a 3

4 particular cloud type. It appears that NWP models will need to employ cloud type specific PDFs formulations in order to achieve the uniform level of errors in precipitation forecast. Figure 2. Comparison of errors in autoconversion (solid lines) and accretion (dashed) rates when applying PDFs obtained from shallow cumulus RICO (black) and congestus TOGA COARE (red) datasets. [graph: The differences are significant for autoconversion rates, but less so for accretion] We also conducted simulations to evaluate the errors of conversion rates in different domain sizes. Figure 3 compares these errors in domain sizes varying from 6.4 to 51.2 km (panels from bottom to top in Fig.3). Color curves refer to PDFs derived using datasets from domains with horizontal grid dimensions from 64 to 512 grid points. Results demonstrate that errors in autoconversion rates have a relatively small sensitivity to the dataset from which the PDF was obtained and this conclusion applies for all domain sizes (from small at the bottom to large at the top panels). However, the accretion rates (shown in the right panels) have more sensitivity to the way PDF is derived. This is especially true for smallest grid resolution of 6.4 km. We conclude that a universal formulation of PDFs can be employed for a range of grid resolutions from 12 to 50 km, however, for the smallest grid resolution of 6.4 km a separately tuned PDF is required. 4

5 Figure 3. Comparison of errors in autoconversion (left panels) and accretion (right panels) rates by applying PDFs obtained from different domain sizes. Panels labeled Dom64, Dom128, Dom256, Dom512 denote domain sizes from 6.4, 12.8, 25.6, and 51.2 km, respectively. Each panel shows color curves of cumulative probability distribution function of errors when using PDF obtained from different domain sizes [graph: The differences are small for autoconversion rates, but for accretion evaluation in the smallest domain the tuning of PDF is required] 5

6 IMPACT The improved parameterization of the physical processes in shallow convective cloud systems will lead to more accurate numerical weather predictions for Navy operations. TRANSITIONS Our results have been published in a refereed scientific paper and reported at the scientific meeting and International Conference on Clouds and Precipitation (Boston, 2014). PUBLICATIONS Kogan, Y. L., D. B. Mechem, 2014: A PDF based microphysics parameterization for shallow cumulus clouds. J. Atmos. Sci., 71, [refereed] Kogan, Y. L., 2014: Parameterization of microphysics in cumulus convective clouds. 14th Conference on Cloud Physics, July 7-11, Boston, MA Kogan, Y. L., 2014: LES based approach to parameterization of cumulus convective cloud systems, Scientific Workshop on ONR DRI Unified Parameterization for Extended Range Prediction, NRL-Monterey, August,